synthetic-data-generate-eval

What problem does it solve?

Manually validating synthetic data generation runs to ensure they produce demo-quality, believable data is time-consuming and prone to oversight, especially when schema-conformant data can still be obviously synthetic and unusable for stakeholder demos.

Core Features & Use Cases

• Multi-dimension grading: Evaluates 6 weighted dimensions including record count health, form schema coverage, out-of-chain data plausibility, warning honesty, manifest provenance, and operator next steps.
• Hard fitness gates: Automatically fails runs with obviously synthetic/garbage data even if they meet all schema conformance requirements, ensuring only believable data passes validation.
• Use Case: For ACE Phase 7 Plan B workflows, this skill validates that labs-side synthetic generation for Connect opportunities produces enough visits, correctly round-trips named FLWs with their promised archetypes, and generates data that would pass a domain expert's casual review of the labs dashboard.

Quick Start

Use the synthetic-data-generate-eval skill to evaluate the latest synthetic data generation run for your active Connect opportunity and produce a formal quality verdict file.